Can ChatGPT relate to you? Exploring consumer satisfaction with AI-generated product advice through the lens of consumption values

As generative AI (GAI) tools increasingly assist consumers in making purchasing decisions, understanding how users evaluate such recommendations has become a timely and essential research question. This study applies the Theory of Consumption Values to examine how consumers develop satisfaction with ChatGPT's product recommendations, while also critically assessing the relevance of each value dimension in the context of AI-mediated interactions. Drawing on a randomized online experiment with U.S. participants, we investigate the relationships between customer satisfaction with ChatGPT's responses and five perceived consumption values: functional, emotional, epistemic, social, and conditional. We also investigate if trust plays a mediating role in these connections and whether they differ depending on the task type (requiring mostly human or machine skills) and product attribute (functional, experiential, hybrid). The findings indicate that, except for social value, all value dimensions are strongly linked to greater levels of satisfaction; the most significant correlation is found for functional value. Trust significantly mediates the value–satisfaction relationship, with conditional value exhibiting the most substantial indirect effect through trust. Interestingly, functional value plays a more prominent role in associating with satisfaction for experiential and hybrid products, as well as in human-skill-intensive tasks, rather than in functional products or machine-skill-oriented tasks. Conditional value consistently shows both direct and mediated associations with satisfaction across functional and hybrid products, as well as both task types. Epistemic value demonstrates both direct and mediated associations with satisfaction in functional and experiential product contexts, as well as in tasks requiring primarily machine-based skills. These findings advance theoretical understanding of value perception in GAI interactions and offer practical insights for designing context-sensitive and trust-oriented recommendation systems.